AI Paper Atlas

Research map / Optimization / Implicit bias and generalization

Towards Understanding the Dynamics of Low-Rank Adaptation

Shu Ding, Yang Peng, Hangan Zhou, Xinyu Lu, Shangwei Chen, Junhua Huang, Mingxuan Yuan, Wei Wang

ICML 2026 · Optimization › Implicit bias and generalization · approach: Low-rank and single-index recovery

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
initialize LoRA A as ETF instead of Gaussian (ablation)LoRA · GLUE CoLA, RoBERTa rank 8accuracy: 57.95 -> 62.18Improved
ETF init instead of Gaussian (ablation)ReLoRA · GLUE MNLI, RoBERTa rank 8accuracy: 82.79 -> 85.16Improved
ETF init instead of Gaussian (ablation)FLORA · GLUE RTE, RoBERTa rank 8accuracy: 72.08 -> 74.85Improved
ETF init instead of default init (ablation)LoRA-GA and PiSSA · GLUE CoLA, RoBERTa rank 8accuracy: 58.70 -> 62.18 (LoRA-GA); 58.14 -> 62.18 (PiSSA)Improved
ETF init instead of Gaussian (ablation)LoRA · GSM8K, Llama-3.1-8B rank 8accuracy: 68.33 -> 71.07Improved
ETF init instead of Gaussian (ablation)LoRA · HumanEval, Llama-3.1-8B rank 8score: 42.48 -> 46.34Improved
ETF init instead of Gaussian (ablation)LoRA · MT-Bench, Mistral-7B-v0.1 rank 8score: 6.13 -> 6.53Improved
ETF init instead of SVD init (ablation)GaLore · GSM8K, Llama-3.1-8B rank 8accuracy: 76.21 -> 78.71Improved
ETF init instead of SVD init (ablation)GaLore · GLUE (7 tasks), RoBERTa rank 4/8/16accuracy: e.g. QQP rank 4: 91.08 -> 90.15; SST-2 rank 8: 95.15 -> 95.12No clear effect
ETF init instead of default init (ablation)LoRA variants · CoLA, MRPC, QNLI, RTE, RoBERTa rank 8training lossImproved
ETF init vs Gaussian and orthogonal init for gradient energy preservation (ablation)rank-4 A on random 100x100 gradient · synthetic gradient, 100 trialsenergy preservation ratioImproved
ETF init across ranks 4, 8, 16 (ablation)LoRA variants · RoBERTa and Llama-3.1-8BaccuracyImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy